AI Research Research Articles

Making Human Learning Visible in a World of Invisible Generative AI

The growth of Generative Artificial Intelligence (GenAI) in Higher Education (HE) challenges us to think differently and creatively about learning and assessment in order to prepare graduates for their futures (Dawson, 2023) This presentation will argue that the very strength of GenAI is also its Achilles’ heel. Anyone can access, garner and generate data rapidly, but human beings have never had a greater need for scrutiny and evaluation of this output. Our graduates will need to have a transparent understanding of how they are learning in terms of the critical choices they make, the information or knowledge they engage with, and where machine learning ends and theirs begins (Okolie et al, 2022). The future of assessment in HE must go beyond simply measuring knowledge acquisition and instead focus on capturing and evidencing students’ capacity for critical thinking. We will highlight some preliminary findings and early impacts of work underway at the University of Bath (UK) and University of Stellenbosch (South Africa), who are working collaboratively to ensure students develop holistically, excel in critical thinking, and gain the confidence and skills to use GenAI effectively to advance their ideas and learning (Osborne, 2024). This will be structured in three parts: Part one (Reviewing): We will outline work during academic year 23/24 by academic staff to review all their assessments against a three-fold categorisation model of permitted GenAI usage. From September 2024, all assessments clearly signal to students the extent to which GenAI is permitted in each assignment. Part two (Guidance): We will showcase resources that support staff to rethink assessment design that embeds GenAI in ways that are authentic, valid and robust (Upsher et al, 2024). We will also present resources that support students to consider how they might use GenAI ethically, and to develop their AI literacy skills and critical thinking. Part three (Examples): We will present initial findings from an internal project that is engaging with national and international employers, in-house world-leading AI researchers and our own students that highlight creative approaches to harnessing GenAI as a tool to reinforce and support (rather than replace) human learning. We conclude by suggesting that HE has always promoted critical thinking and the rigorous investigation and examination of the production of knowledge. We need graduates who can critically engage with their learning, and assessments that capture and evidence this. The only way to do this effectively is to put people at the start, middle and end of our assessment journey.

Smart Caption: Intelligent Image Description Using Transformer Decoder

The need for automated image description systems has grown significantly with the rise of multimedia content across diverse domains such as social media, digital libraries, and accessibility technologies. Smart Caption presents a novel approach for generating intelligent and context-aware image descriptions by utilizing a combination of Vision Transformers (ViTs) and Natural Language Processing (NLP) Transformer decoders. Unlike traditional convolution-based methods, Vision Transformers treat images as sequences of patches, enabling them to capture global image features more effectively. These extracted visual features are then processed by a Transformer-based decoder to produce coherent and contextually appropriate captions, bridging the gap between image recognition and natural language generation. Our approach focuses on leveraging the attention mechanisms inherent in both Vision and NLP Transformers to enhance the quality of image descriptions. The ViT architecture is designed to focus on relevant regions within an image, while the NLP Transformer decoder is adept at generating fluent and detailed descriptions by attending to the most significant visual features. This two-stage process improves the precision of object detection, relationship understanding, and scene context in generated captions. Experimental results demonstrate that Smart Caption out performs traditional methods in terms of accuracy, contextual relevance, and fluency, marking a significant step forward in the field of automated image captioning. Through this research, we aim to provide a scalable, efficient, and intelligent solution for image description, with potential applications in areas such as accessibility tools for visually impaired users, digital content indexing, and automated content generation. Our findings highlight the strengths of transformer-based models in bridging vision and language tasks, offering promising directions for future research in multimodal AI..

StructuGraphRAG: Structured Document-Informed Knowledge Graphs for Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating external data sources beyond their training sets and querying predefined knowledge bases to generate accurate, context-rich responses. Most RAG implementations use vector similarity searches, but the effectiveness of this approach and the representation of knowledge bases remain underexplored. Emerging research suggests knowledge graphs as a promising solution. Therefore, this paper presents StructuGraphRAG, which leverages document structures to inform the extraction process and constructs knowledge graphs to enhance RAG for social science research, specifically using NSDUH datasets. Our method parses document structures to extract entities and relationships, constructing comprehensive and relevant knowledge graphs. Experimental results show that StructuGraphRAG outperforms traditional RAG methods in accuracy, comprehensiveness, and contextual relevance. This approach provides a robust tool for social science researchers, facilitating precise analysis of social determinants of health and justice, and underscores the potential of structured document-informed knowledge graph construction in AI and social science research.

AI and dental research

AI and dental research

Building Trustworthy AI: The Role of Patient and Public Involvement in Healthcare AI Development

AI is helping researchers make great strides in healthcare. However, there is a trust deficit in AI when applied to critical areas like healthcare. Hence communicating the beneficial medical applications of AI and engaging the public with healthcare AI research is critical. One way to achieve this is by getting the community involved in co-designing better AI systems in healthcare projects. We argue that AI algorithms for healthcare should be co-designed with patients and healthcare workers, so that they are useful and trustworthy. We suggest a roadmap for this collaborative approach in AI model building. This will involve actively including patients with lived experience of a disease, as well as creating a research advisory group to walk patients through the process of AI model building. We suggest formulating and scoping a problem, and then generating a hypothesis that patients and scientists agree on. The road to building trustworthy AI systems may become easier if all stakeholders are involved in co-creating AI models.

RAI Guidelines: Method for Generating Responsible AI Guidelines Grounded in Regulations and Usable by (Non-)Technical Roles

Many guidelines for responsible AI have been suggested to help AI practitioners in the development of ethical and responsible AI systems. However, these guidelines are often neither grounded in regulation nor usable by different roles, from developers to decision makers. To bridge this gap, we developed a four-step method to generate a list of responsible AI guidelines; these steps are: (1) manual coding of 17 papers on responsible AI; (2) compiling an initial catalog of responsible AI guidelines; (3) refining the catalog through interviews and expert panels; and (4) finalizing the catalog. To evaluate the resulting 22 guidelines, we incorporated them into an interactive tool and assessed them in a user study with 14 AI researchers, engineers, designers, and managers from a large technology company. Through interviews with these practitioners, we found that the guidelines were grounded in current regulations and usable across roles, encouraging self-reflection on ethical considerations at early stages of development. This significantly contributes to the concept of 'Responsible AI by Design'- a design-first approach that embeds responsible AI values throughout the development lifecycle and across various business roles.

Making sense of generative AI for assessments: Contrasting student claims and assessor evaluations

The rapid growth of generative AI usage in higher education has left educators looking urgently for insights into student usage and guidance on good practice. This case study examines an experiential exercise involving 118 postgraduate management students at a UK business school, where students were asked to write a 500-word reflection on their use of AI for a 2500-word essay-style assessment. Using sensemaking as a theoretical lens, we compare students' claims with assessors' evaluations of students' AI usage. Our findings indicate that students predominantly use generative AI for writing, paraphrasing, and rephrasing, rather than for fostering critical thinking or engaging in the more advanced stages of sensemaking, a level achieved by only one-tenth of the cohort. The consistency between this study's findings and pre-generative AI research suggests that higher education has yet to adapt adequately in ways to integrate AI to mitigate, rather than exacerbate, current sector deficiencies. We call on university leaders to develop institutional strategies that allow for effective and responsible integration of generative AI, and on educators to develop students' critical evaluation and academic writing skills that build on generative AI's affordances, with several specific recommendations made in this article.

Articulation Work and Tinkering for Fairness in Machine Learning

The field of fair AI aims to counter biased algorithms through computational modelling. However, it faces increasing criticism for perpetuating the use of overly technical and reductionist methods. As a result, novel approaches appear in the field to address more socially-oriented and interdisciplinary (SOI) perspectives on fair AI. In this paper, we take this dynamic as the starting point to study the tension between computer science (CS) and SOI research. By drawing on STS and CSCW theory, we position fair AI research as a matter of 'organizational alignment': what makes research 'doable' is the successful alignment of three levels of work organization (the social world, the laboratory, and the experiment). Based on qualitative interviews with CS researchers, we analyze the tasks, resources, and actors required for doable research in the case of fair AI. We find that CS researchers engage with SOI research to some extent, but organizational conditions, articulation work, and ambiguities of the social world constrain the doability of SOI research for them. Based on our findings, we identify and discuss problems for aligning CS and SOI as fair AI continues to evolve.

AI in business research

AbstractArtificial intelligence (AI) has emerged as a pivotal force in modern business transformation, garnering widespread attention from both practitioners and academics. With a notable exponential increase in AI‐related studies, we provide a research framework aiming to synthesize the existing literature on AI in the business field. We conduct a comprehensive review of AI research spanning from 2010 to 2023 in 25 leading business journals according to this review framework. Specifically, we review the literature from three research perspectives: (i) AI applications, (ii) human perceptions of AI, and (iii) AI behavior. We also identify five principal research questions and offer suggestions for future research directions.

Exploring Core Knowledge in Interdisciplinary Research: Insights from Topic Modeling Analysis

Although interdisciplinary research has garnered extensive attention in academia, its core knowledge structure has yet to be systematically explored. To address this gap, this study aims to uncover the underlying core knowledge topics within interdisciplinary research, enabling researchers to gain a deeper understanding of the knowledge framework, improve research efficiency, and offer insights for future inquiries. Based on the Web of Science (WoS) database, this study collected 153 highly cited papers and employed the LDA topic model to identify latent topics and extract the knowledge structure within interdisciplinary research. The findings indicate that the core knowledge topics of interdisciplinary research can be categorized into four major areas: the knowledge framework and social impact of interdisciplinary research, multidisciplinary approaches in cancer treatment and patient care, Covid-19 multidisciplinary care and rehabilitation, and multidisciplinary AI and optimization in industrial applications. Moreover, the study reveals that AI-related interdisciplinary research topics are rapidly emerging. Through an in-depth analysis of these topics, the study discusses potential future directions for interdisciplinary research, including the cultivation and development of interdisciplinary talent, evaluation systems and policy support for interdisciplinary research, international cooperation and interdisciplinary globalization, and AI and interdisciplinary research optimization. This study not only uncovers the core knowledge structure of interdisciplinary research but also demonstrates the effectiveness of the LDA topic model as a data mining tool for revealing key topics and trends, providing practical tools for future research. However, this study has two main limitations: the time lag of highly cited papers and the dynamic evolution of interdisciplinary research. Future research should address these limitations to further enhance the understanding of interdisciplinary research.

AI and Research: Methods and Best Practices for Conducting Research Using AI

This video presentation provides an overview of methods and best practices for using AI in academic research, emphasizing its significance from a library perspective. It introduces key concepts, strategies, and the importance of AI literacy, incorporating a brief review of frameworks for the evaluation and use of AI technologies. This video also offers a concise summary of popular AI research tools and best practices based on select recommendations from reputable institutions and industry scholarship. The inclusion of visual and auditory elements detailing key matters pertaining to the use of AI in higher education contributes to a multimodal and diversified approach designed for a broader audience. Video formats may complement other traditional forms of publication by providing visual explanations that are intended to enhance accessibility and engagement with complex concepts in scholarly discourse.

The Research Advance in Ethical Issues of AI Application

The topic of AI has become more and more discussed ever since the release of ChatGPT, as its astonishing intelligence has made humans to challenge the boundary. This vagueness then contributes to the emergence of discussions on the ethics of AI. Aiming to clarify the many current academic discussions on the ethics of artificial intelligence and to recognize different ethical perspectives, this study conducts a literature review of the research and discussions among different scholars on ethical issues in the application of artificial intelligence technologies. It is found that the current relevant discussion extends from a micro perspective emphasizing the individual to a macro perspective emphasizing the human race. Research advances on ethical issues in AI applications can be categorized into three levels: personal, social, and global. Among which, the individual level focuses on the ethical issues in algorithmic logic and user data management. The social level focuses on the social employment and public accountability systems. While the global level discusses intellectual property disputes and the social identity of AI robots. Finally, the author gave an evaluation of the gaps and challenges of studying AI ethics and emphasize the importance of addressing ethical issues in AI development and deployment. The paper aims to serve as information for scholars and policymakers, fostering ethical awareness in the AI research community and helping AI technologies to be developed and used responsibly.

Use of generative AI in research: ethical considerations and emotional experiences

ABSTRACT This study examines researchers’ ethical concerns toward the deployment of GenAI in research and their emotional responses. To acquire an in-depth understanding, we used narrative frames and follow-up interviews to collect data from 22 researchers who reported extensive experience with GenAI. An inductive thematic analysis revealed three themes capturing ethical concerns that invoked three types of emotional reactions. From an ethical perspective, our participants were concerned with “human ethical agency in AI research practices,” “cognitive impacts of overreliance on GenAI in research,” and “ethical issues of access, accuracy, and privacy.” From an emotional perspective, they showed “mixed emotions,” “positive emotions,” and “negative emotions” when dealing with GenAI tools. There were close connections between the ethical implications of GenAI and the emotional reactions to them. In this light, we conclude that emotional reactions to GenAI, which are determinants of future use, should be taken more seriously in further research.

Towards Self-Conscious AI Using Deep ImageNet Models: Application for Blood Cell Classification

The exceptional performance of ImageNet competition winners in image classification has led AI researchers to repurpose these models for a whole range of tasks using transfer learning (TL). TL has been hailed for boosting performance, shortening learning time and reducing computational effort. Despite these benefits, issues such as data sparsity and the misrepresentation of classes can diminish these gains, occasionally leading to misleading TL accuracy scores. This research explores the innovative concept of endowing ImageNet models with a self-awareness that enables them to recognize their own accumulated knowledge and experience. Such self-awareness is expected to improve their adaptability in various domains. We conduct a case study using two different datasets, PBC and BCCD, which focus on blood cell classification. The PBC dataset provides high-resolution images with abundant data, while the BCCD dataset is hindered by limited data and inferior image quality. To compensate for these discrepancies, we use data augmentation for BCCD and undersampling for both datasets to achieve balance. Subsequent pre-processing generates datasets of different size and quality, all geared towards blood cell classification. We extend conventional evaluation tools with novel metrics—“accuracy difference” and “loss difference”—to detect overfitting or underfitting and evaluate their utility as potential indicators for learning behavior and promoting the self-confidence of ImageNet models. Our results show that these metrics effectively track learning progress and improve the reliability and overall performance of ImageNet models in new applications. This study highlights the transformative potential of turning ImageNet models into self-aware entities that significantly improve their robustness and efficiency in various AI tasks. This groundbreaking approach opens new perspectives for increasing the effectiveness of transfer learning in real-world AI implementations.

The Framework of GeoSOT-3D Grid Modeling for Spatial Artificial Intelligence

Abstract. With the development of society, spatial artificial intelligence (spatial AI) research is gradually able to play a greater role. However, spatial AI has problems such as data alignment, poor interpretability, and cross domain learning. Therefore, this paper proposes an innovative GeoSOT-3D grid modeling framework for spatial AI research, which enhances the application capabilities of spatial AI. Grid modeling will be able to run through the upstream and downstream of spatial AI research, providing encoding calculations and spatial neighborhood embedding matrices for spatial data. This paper also uses task examples to demonstrate how to effectively organize and index spatial data using GeoSOT-3D grids and conduct spatial AI research. The use of GeoSOT-3D grids for spatial AI analysis has enormous potential and broad application prospects, which will help promote the further development and application of spatial AI.

Biofeedback-Based Closed-Loop Phytoactuation in Vertical Farming and Controlled-Environment Agriculture.

This work focuses on biohybrid systems-plants with biosensors and actuating mechanisms that enhance the ability of biological organisms to control environmental parameters, to optimize growth conditions or to cope with stress factors. Biofeedback-based phytoactuation represents the next step of development in hydroponics, vertical farming and controlled-environment agriculture. The sensing part of the discussed approach uses (electro)physiological sensors. The hydrodynamics of fluid transport systems, estimated electrochemically, is compared with sap flow data provided by heat-based methods. In vivo impedance spectroscopy enables the discrimination of water, nutrient and photosynthates in the plant stem. Additionally to plant physiology, the system measures several air/soil and environmental parameters. The actuating part includes a multi-channel power module to control phytolight, irrigation, fertilization and air/water preparation. We demonstrate several tested in situ applications of a closed-loop control based on real-time biofeedback. In vertical farming, this is used to optimize energy and water consumption, reduce growth time and detect stress. Biofeedback was able to reduce the microgreen production cycle from 7 days to 4-5 days and the production of wheatgrass from 10 days to 7-8 days, and, in combination with biofeedback-based irrigation, a 30% increase in pea biomass was achieved. Its energy optimization can reach 25-30%. In environmental monitoring, the system performs the biological monitoring of environmental pollution (a low concentration of O3) with tomato and tobacco plants. In AI research, a complex exploration of biological organisms, and in particular the adaptation mechanisms of circadian clocks to changing environments, has been shown. This paper introduces a phytosensor system, describes its electrochemical measurements and discusses its tested applications.

Dental Implant Identification Methods

Dental implants are a well-accepted prosthetic alternative for missing teeth. After implant restoration, they will need maintenance in due course of time due to biological and mechanical complications, during which information about the implant system is required. Until today there is no well-established method for implant identification and conventional tools such as interpretation from radiographs need time and effort. Researchers have proposed several methods for implant identification and the review focuses on a comprehensive discussion of the proposed methods. For this review, comprehensive data from databases, including PubMed, Scopus, Web of Science, Cochrane, and Google Scholar, was thoroughly examined ensuring the most up-to-date and relevant information regarding implant identification. The proposed methods include an interpretation from radiographs based on the implant design specifications listed, implant records, implant recognition software, retrieving implant information through a wireless reader from a radiofrequency chip fitted into an implant screw hole, QR-encoded implant identification wallet, bar code encryption by implant manufacturers, incorporating laser-etched batch and serial numbers in implant collars, Sharma Jhingta system of implant identification and artificial intelligence methods. Amongst existing methods, AI research shows potential in offering a quick and accurate method of implant identification however developing a robust AI model with a comprehensive database is a complex task and requires considerable effort and time.

Prompting Considered Harmful

As systems graduate from labs to the open world, moving beyond prompting is central to ensuring that AI is useful, usable, and safe for end users as well as experts such as AI developers and researchers.

Intelligent Vision System with Pruning and Web Interface for Real-Time Defect Detection on African Plum Surfaces

Agriculture stands as the cornerstone of Africa’s economy, supporting over 60% of the continent’s labor force. Despite its significance, the quality assessment of agricultural products remains a challenging task, particularly at a large scale, consuming valuable time and resources. The African plum is an agricultural fruit that is widely consumed across West and Central Africa but remains underrepresented in AI research. In this paper, we collected a dataset of 2892 African plum samples from fields in Cameroon representing the first dataset of its kind for training AI models. The dataset contains images of plums annotated with quality grades. We then trained and evaluated various state-of-the-art object detection and image classification models, including YOLOv5, YOLOv8, YOLOv9, Fast R-CNN, Mask R-CNN, VGG-16, DenseNet-121, MobileNet, and ResNet, on this African plum dataset. Our experimentation resulted in mean average precision scores ranging from 88.2% to 89.9% and accuracies between 86% and 91% for the object detection models and the classification models, respectively. We then performed model pruning to reduce model sizes while preserving performance, achieving up to 93.6% mean average precision and 99.09% accuracy after pruning YOLOv5, YOLOv8 and ResNet by 10–30%. We deployed the high-performing YOLOv8 system in a web application, offering an accessible AI-based quality assessment tool tailored for African plums. To the best of our knowledge, this represents the first such solution for assessing this underrepresented fruit, empowering farmers with efficient tools. Our approach integrates agriculture and AI to fill a key gap.

Explaining AI through mechanistic interpretability

Recent work in explainable artificial intelligence (XAI) attempts to render opaque AI systems understandable through a divide-and-conquer strategy. However, this fails to illuminate how trained AI systems work as a whole. Precisely this kind of functional understanding is needed, though, to satisfy important societal desiderata such as safety. To remedy this situation, we argue, AI researchers should seek mechanistic interpretability, viz. apply coordinated discovery strategies familiar from the life sciences to uncover the functional organisation of complex AI systems. Additionally, theorists should accommodate for the unique costs and benefits of such strategies in their portrayals of XAI research.